Identification And Verification Of Key Biomarkers And Potential Therapeutic Targets In Diabetic Kidney Disease Using Multi-omics Data

Background: Diabetic kidney disease(DKD)is a microvascular complication of diabetes,occurs in about 20～40% of diabetic patients.In order to improve the prognosis of DKD patients,on the one hand,blood/urine non-invasive biomarkers can be used to identify high-risk patients for early intervention.On the other hand,existing treatment options can be improved by developing new drug targets.Estimated glomerular filtration rate(e GFR)and albuminuria are the most commonly used clinical biomarkers,however,they lack sensitivity and specificity in predicting progressive renal decline and end-stage renal disease(ESRD).The traditional therapeutic regimen of rigorous blood sugar control,application of angiotensin converting enzyme inhibitor(ACEI)or angiotensin II receptor blocker(ARB)to control blood pressure is not enough to prevent or reverse DKD progression.Therefore,there is an urgent need to develop novel biomarkers and drug targets to delay the progression of DKD.In the past decades,with the development of highthroughput omics technology and its application in population-based cohort,our understanding of molecular mechanism of DKD has much improved.In a step further,the combination of high-throughput omics data and phenotype data could facilitate the identification of DKD prognosis biomarker as well as potential therapeutic target.Objective: To improve the prognosis of DKD,this research aims to:1.Identify and verify prognostic markers of DKD using plasma proteome data from prospective cohort.2.Integrate plasma proteome with kidney transcriptome data to reveal the molecular mechanism underlying the association between key biomarkers and DKD progression.3.Construct robust gene co-expression network through in-depth analysis of transcriptome data from renal tubulointerstitium,and screen potential therapeutic target based on the network.Methods: 1.Multi-scalar data integration to identify and validate DKD prognostic biomarker:1)Plasma samples from two groups of patients were selected from the Clinical Phenotyping and Resource Biobank Core(C-PROBE)of the University of Michigan to construct a discovery cohort and a validation cohort.With the aid of the nucleic acid aptamer-based proteomics tool SOMAScan,1305 plasma protein expressions at the baseline level of the enrolled patients were detected.In the discovery cohort,univariate Cox proportional hazards models were used to screen for proteins associated with DKD prognosis,where the primary end point event was defined as achievement of ESRD or a 40% reduction in e GFR compared to baseline.Further,using the machine learning method Lasso Cox,through repeated k cross-validation(5 fold,200 repeat),a multi-biomarker joint prediction model that minimizes the cross-validation error is obtained,and its improvement on top of existing clinical models is evaluated.2)For the multi-biomarker model constructed in the discovery cohort,C statistic was used to evaluate its predictive effect on the progression of DKD in the validation cohort.For the key biomarkers in the model,Angiopoietin-2(ANGPT2),the SOMAScan proteomics results were further validated by ELISA.3)Using data from three signaling pathway databases(Reactome,NETPATH,PID),the ANG-TIE signaling pathway gene set downstream of ANGPT2 was constructed.Further,based on the glomerular and tubulointerstitial transcriptomic data from C-PROBE,the z-score method was used to calculate the ANG-TIE pathway activation score in the glomerulus and tubulointerstitium.Finally,the proteomic and transcriptomic data were integrated to explore the association of ANGTIE pathway activation fraction with plasma ANGPT2 concentration and DKD prognosis,and validated in external data.4)Integrate the kidney transcriptome data from C-PROBE and the single-cell transcriptome data from the Kidney Precision Medicine Project(KPMP)to explore the distribution of TIE2/TEK receptors downstream of ANGPT2 in kidney cells and ANG-TIE activation of signaling pathways in different kidney cells.Finally,the differences of TIE2/TEK receptor and ANG-TIE signaling pathway between DKD and normal controls were evaluated.2.Integrate bulk and single-cell transcriptome data to explore potential therapeutic targets of DKD.1)Search the public DKD tubulointerstitial transcriptome data from the GEO and Pubmed database,and use the consensus WGCNA(weighted gene co-expression network analysis)method to construct a multi-dataset conserved gene co-expression network/module.Further screen for DKD-related modules,functional annotation,network visualization,and assessment of network conservation in external validation data.2)Download the single-cell nuclear transcriptome data containing 3DKD kidney samples from the GEO database for re-analysis.Explore the cellular localization of the conserved DKD-related gene co-expression network in the kidney.3)Use the CMap tool to find compounds or genetic interventions that may reverse the co-expression network.Transcription factors in potential regulatory networks were searched using i Regulon.For the top transcription factors,we further used single-cell nuclear transcriptome data to determine their cellular localization,and combined with Nephroseq data to assess their association with renal function.Results: 1.Identification and validation of DKD prognostic biomarkers:1)Using the SOMAScan proteomic data of the discovery cohort(n=58),84 plasma proteins(P<0.05)associated with the prognosis of DKD were identified,and a three-biomarker combined prediction model(ANGPT2,CLEC4 M and EGFR)was generated.This model improved the C statistic of the clinical model from 0.728 to 0.791(Likelihood ratio test P value=0.003)and was validated in the validation cohort(n=68)(Likelihood ratio test P value=0.0004).2)The biomarker ANGPT2 was significantly associated with DKD prognosis in the discovery cohort and validation cohort,even after adjusting for age,sex,race,e GFR,and the biomarkers EGFR and CLEC4M(discovery cohort hazard ratio [Hazard Ratio,HR] = 3.39,P=0.017;validation cohort HR=3.59,P=0.006).ANGPT2 level measured by ELISA was highly correlated with that detected by SOMAScan(r=0.91,P=2.58e-23).3)A gene set of ANG-TIE signaling pathway consisting of 154 genes was constructed.ANG-TIE signaling pathway score was significantly associated with plasma ANGPT2 protein(r=0.43,P=0.01)and DKD prognosis(P=0.02),but not in tubulointerstitium.This phenomenon can be partly explained by the significantly higher expression of the ANGPT2 receptor TEK in glomerulus than tubulointerstitium.2.Explore potential therapeutic targets for DKD based on public data: Two conserved DKD-related gene co-expression modules(M2,M3)were mined.M2 is an inflammation-and fibrosis-related network that is up-regulated in DKD.It is active in immune cells,fibroblasts,vascular smooth muscle cells,and peritubular endothelial cells,and is regulated by the transcription factor NF-κB.Another network,M3,is a metabolismrelated network that is downregulated in DKD.It is active in proximal tubule cells,and is regulated by transcription factors such as HNF4 A and ESRRA,and can be reversed by HDAC inhibitors.Conclusion: 1.We identified a series of plasma proteins associated with DKD progression,and constructed a multi-biomarker model for predicting DKD progression.We further revealed the link between plasma ANGPT2,glomerular ANG-TIE signaling and DKD prognosis,which support the targeted therapy of ANG-TIE signaling pathway.2.Important DKD-related gene co-expression networks were discovered,and important network regulators(HNF4A,ESRRA and HDAC inhibitors,etc.)were screened,providing a general framework guiding further experimental work.